E-commerce Pilot

Read Amazon store insights, upload reports, and propose ad changes that wait for approval.

E-commerce Pilot turns a seller's own Amazon data into sales, profit, stock and advertising insights, and runs seven strategies over it that find wasted ad spend and suggest better bids. The API gives your code, or an AI assistant you already use, the same data and most of the same controls as the dashboard. Every write goes through the same checks the dashboard applies, so nothing reaches Amazon that a person in the workspace has not allowed. Amazon is the only channel today; Shopify, eBay and TikTok Shop are coming soon.

Scopes

Scopes match the service permissions, so they can be granted to custom roles as well as to keys.

• service.ecommerce-pilot.view - read sales, profit, stock, ads, strategies, findings, keywords, listings, research, reimbursements and alerts, preview rules, and mark alerts read.
• service.ecommerce-pilot.upload - upload Amazon report files for a store.
• service.ecommerce-pilot.sync - start a sync from a connected Amazon store.
• service.ecommerce-pilot.settings - add stores that work from uploaded reports, rename a store or record Brand Registry, set the target ACOS, thresholds and product costs, and update reimbursement cases.
• service.ecommerce-pilot.ads.manage - propose, apply and undo bid, budget, negative keyword and campaign changes, dismiss or snooze findings, and write analysis for findings.
• service.ecommerce-pilot.listings.manage - save, check and submit listing drafts, and send review requests.
• service.ecommerce-pilot.research - run product hunting, track competitor products, keep keyword lists and save research written by your own model.
• service.ecommerce-pilot.export - download store data and strategies as CSV.

Creating, changing or deleting an automatic rule needs both service.ecommerce-pilot.ads.manage and service.ecommerce-pilot.settings, the same pair the dashboard asks for.

What the API can do

• Read insights. Overview figures for a date range, profit by product or by day, stock with reorder dates, campaigns, keywords and search terms with their metrics, orders by weekday and hour, each of the seven strategies, the findings they raised, listings with a health score, reimbursement cases, alerts and sync status. GET /guide returns the rules, formulas and thresholds the dashboard reasons with, so your model works from the same definitions.
• Stores and uploads. Add a store that works from uploaded reports and send report text for it. The report type is recognised from its columns, the file is imported in the background, and ecommerce-pilot.upload.completed fires when it is done. Buyer details in an order file are dropped on import. Connecting a store to Amazon is done by a person in the dashboard.
• Sync. Start a sync from a connected store and follow it through the task events.
• Propose and apply changes. Bid, budget and negative keyword changes, pausing or enabling a campaign, applying the change a finding suggests, and undoing an applied change.
• Listings. Save a listing draft your model wrote, check it against the listing rules and Amazon, and submit it. Send review requests for orders that can still get one.
• Automatic rules. List, create, change, delete and preview rules. Preview shows what a rule would change now, without changing anything.
• Research.Run product hunting (Amazon catalogue, pricing and fee lookups, scored for opportunity, with sales estimated from rank), keep keyword lists, and add or remove the competitor products you watch. Keyword ideas come from the store's own data and Amazon.
• Save your own model's analysis. Write the explanation for a finding, advice for a keyword in a list, or a research report. It is stored with the name of the app that wrote it and shown in the dashboard with a Written by badge.
• Settings and costs. Read the settings, set the target ACOS and the strategy thresholds, and set product costs.
• Export. Download a strategy or store data as CSV.

Dashboard-only settings

The guardrails, whether changes from AI apps and API keys go to review or apply directly, whether rules may apply changes automatically, and how long data is kept can only be changed by a person in the dashboard. Sending any of them to PUT /settings is refused with 403 and the code setting_dashboard_only. The daily summary email is also switched on and off in the dashboard.

Changes wait for approval

By default a change from an API key or a connected AI app is not sent to Amazon straight away. It goes to AI Review in the workspace, and the response carries status: "pending_approval" with the id of the review, so you can tell the user it is waiting. Once someone with permission approves it, it is applied and ecommerce-pilot.change.applied fires. Undo works the same way: while the workspace sends changes to review, an undo from the API waits there too.

A workspace can choose in the dashboard to let changes from AI apps and API keys apply directly. Even then, a change that moves a bid or raises a budget further in one step than the guardrails allow still goes to AI Review.

A store that was added by uploading reports is not connected to Amazon, so its applied changes become a to-do list for the seller rather than being sent, and the response says so.

Rules while changes need review

While the workspace sends changes from AI apps and API keys to review, an API key or AI app cannot switch a rule on, change a rule that is already on, or change the target ACOS or thresholds while any rule is on. Those calls are refused with 403 and setting_dashboard_only. It can still create rules switched off and preview them, so a person can review and switch them on in the dashboard.

Plan limits

Each plan caps the number of stores (1, 2, 5 or 15), tracked products (10, 50, 200 or 1,000) and research runs per month (5, 30, 150 or 500), from Free to Business. Product hunting searches count as research runs; research your own model writes and saves does not. Going over a limit answers 402 with quota_exceeded.

Use it from Claude, ChatGPT or any MCP client

The hosted MCP server exposes E-commerce Pilot as tools, filtered to the scopes the member approved when they connected. This is the same Connected AI apps flow every service uses, and nothing extra needs setting up inside E-commerce Pilot: add the MCP address to the assistant, sign in with OAuth, pick the workspace and tick the E-commerce Pilot permissions. See Connect an AI assistant for the full flow. Writes made this way are recorded in the audit log against the member who connected the app.

Webhooks

• ecommerce-pilot.sync.completed and ecommerce-pilot.sync.failed - a store sync finished or failed.
• ecommerce-pilot.upload.completed - an uploaded report was imported.
• ecommerce-pilot.finding.created - a strategy found something new.
• ecommerce-pilot.change.applied and ecommerce-pilot.change.undone - a change reached Amazon, or was undone.
• ecommerce-pilot.alert.created - a store alert was raised.
• ecommerce-pilot.task.started, .completed, .failed and .cancelled - background work such as an upload import or a sync.